基于无人机热成像的棉花根域土壤水分含量估测研究

Estimation of soil moisture content in cotton rhizosphere based on UAV thermal imaging

关键词:  棉花  无人机  冠层温度  冠气温差  土壤含水量

Key words:  cotton  UAV  canopy temperature  canopy-air temperature difference  soil moisture content

摘要:

为探究棉花冠层无人机热红外影像对根域土壤水分含量反演规律，以棉花花铃期土壤水分为研究对象，利用小区水分胁迫控制试验和模型模拟的方法，获取棉花花铃期在不同干旱胁迫天数下（灌水后第2、10和32天）的叶片含水量、根域各土层（0~10、10~20、20~30、30~40、40~50 cm）土壤相对含水量（Relative soil water content，RSWC），同步获取无人机热红外影像数据，通过将棉花冠层温度与叶片、大气温度结合构建差值温度指数（Difference temperature index，DTI），比值温度指数（Ratio temperature index，RTI）和归一化差值温度指数（Normalized difference temperature index，NDTI），选择与RSWC相关性在0.6以上的指标（冠层温度、冠-高温差、冠-均温差、DTI2、DTI3、NDTI2、NDTI3）构建不同土层水分含量估测模型，并进行验证。结果表明：1）随干旱胁迫天数的增加，棉花叶片含水量呈先升高后降低的趋势，土壤相对含水量随干旱胁迫天数的增加逐渐降低；2）冠层温度协同叶片、大气温度构建的温度指数估测根域土壤水分含量精度要优于单一使用冠层温度，具体表现为NDTI3>DTI3>DTI2>NDTI2>冠-均温差>冠-高温差>冠层温度；3）多元线性回归模型对棉花根域土壤含水量具有较高的估测精度，其中干旱胁迫第2天对10~20 cm根域土层的估测精度最高，R²为0.600，RMSE为0.388；第10天对20~30 cm根域土层的估测精度最高，R²为0.721，RMSE为0.267；第32天对30~40 cm根域土层的估测精度最高，R²为0.918，RMSE为0.068。因此，利用无人机热成像仪获取棉花冠层温度并结合叶片及大气温度，可对作物根域土壤水分含量进行拟合估测，对发展节水灌溉等土壤水分调控与管理技术具有现实意义。

Abstract:

In order to explore the inversion law of soil moisture content in the cotton rhizosphere based on the thermal infrared UAV imageing， this study took cotton at the flowering and boll stage as research object， and used plot water stress control experiment and model simulation to obtain the leaf water content and the relative soil water content （RSWC） of each soil layer （0-10， 10-20， 20-30， 30-40， 40-50 cm） in the rhizosphere under different drought stress days （2nd， 10th and 32nd days after irrigation） at the flowering and boll stage of cotton， and thermal infrared image data of UAV on the 2nd， 10th and 32nd days of drought stress. The difference temperature index （DTI）， ratio temperature index （RTI）， normalized difference temperature index （NDTI） were constructed by combining cotton canopy temperature with leaf temperature and atmospheric temperature， and the indexes （canopy temperature， canopy-high temperature difference， canopy-average temperature difference， DTI2， DTI3， NDTI2， NDTI3） with a correlation of more than 0.6 with RSWC were selected to fit and verify the moisture content of different soil layers. The results showed that： 1） With the increase of drought stress days， the moisture content of cotton leaves showed a trend of first increasing and then decreasing， and the relative soil moisture content gradually decreased； 2） The precision of the temperature index constructed by canopy temperature with leaf temperature and atmospheric temperature to fit the rhizophere soil water content was better than that of the single use of canopy temperature. The overall trend NDTI3>DTI3>DTI2>NDTI2>canopy-average temperature difference>canopy-high temperature difference>canopy temperature； 3） The multiple linear regression model had a high estimation accuracy for the soil moisture content of cotton rhizophere， and the prediction accuracy of 10-20 cm rhizophere soil layer on the 2nd day of drought stress was the highest （R² is 0.600， RMSE is 0.388）； The prediction accuracy of 20-30 cm rhizophere soil layer on the 10th day is the highest （R² is 0.721， RMSE is 0.267）； and the prediction accuracy of 30-40 cm rhizophere soil layer on the 32nd day is the highest （R² is 0.918， RMSE is 0.068）. In conclusion， the cotton canopy temperature obtained by UAV thermal image combined with leaf temperature and atmospheric temperature can be used to fit and estimate soil moisture content in crop root domain， which has practical significance for the development of water-saving irrigation and other soil water regulation and management technologies.